AccuNGS: detecting ultra-rare variants in viruses from clinical samples

Next generation sequencing is widely used to characterize genetic diversity in a sample, yet is hindered by its relatively low resolution. Particularly, detecting rare genetic variants in clinical samples of viruses is still nearly impossible. Here we describe AccuNGS, an approach that combines error reduction in each sequencing stage with in silico error elimination, which enables detection of variants as rare as 1:10,000 or lower. We thoroughly explore AccuNGS background errors and reveal they are mostly generated in the sequencer itself. We demonstrate that as opposed to common assumptions, Illumina paired-end reads are not independent. After applying AccuNGS to an HIV sample taken during acute infection, we reveal that the vast majority of transition variants in the sample segregate at ultra-low frequencies, rendering them undetectable by standard sequencing. These results highlight the early rich accumulation of genetic diversity during viral infection at depths previously unseen.